nim.py

@@ -31,16 +31,10 @@ class Nim():
 
 
 class NimAI():
-    def __init__(self, alpha=0.5, epsilon=1):
+    def __init__(self, alpha=0.5, epsilon=0.1):
         self.q = dict()  # Q-value table
-        # self.q[(0, 0, 0, 2), (3, 2)] = -1 # Test Q-Value 
-        # self.q[(0, 0, 0, 2), (3, 1)] = 10 # Test Q-Value 
-        
-        self.q[((1,1,1,0), (0,1))] = 0.4
-        self.q[((1,1,1,0), (1,1))] = 0.9
-        self.q[((1,1,1,0), (2,1))] = 0.7
-        self.q[((2,1,1,0), (0,1))] = 0.2
-
+        self.q[(0, 0, 0, 2), (3, 2)] = -1 # Test Q-Value 
+        self.q[(0, 0, 0, 2), (3, 1)] = 10 # Test Q-Value 
         
         self.alpha = alpha  # Learning rate
         self.epsilon = epsilon  # Exploration rate
@@ -95,17 +89,15 @@ class NimAI():
         float: The highest Q-value among available actions. 
                Returns 0 if no actions are available.
     """
-        state = tuple(state)
         # actions = []
-        # for q in self.q:
-        #     print(q)
-        #     if q[0] == tuple(state):
+        # for q in self.q.key:
+        #     if q[0] == state:
         #         actions.append(q[1])
-        actions = [key[1] for key in self.q if key[0] == state]
-        # print(actions)
-        # print(self.q[state, action] for action in actions)
-        # print(max(self.q[state, action] for action in actions))
-        return 0 if actions == [] else max(self.q[state, action] for action in actions)
+        actions = tuple([key[1] for key in self.q.keys() if key[0] == state])
+        try:
+            return max([q for q in self.q[tuple(state), actions]])
+        except:
+            return 0
 
     def choose_action(self, state, epsilon=True):
         """

test.py

@@ -10,17 +10,10 @@ def test_get_q_value(ai):
 
 def test_update_q_value(ai):
     print("\n--- Testing update_q_value ---")
-    state = (2, 1, 1, 0)
-    action = (0, 1)
-    print(ai.q)
-    print(ai.update_q_value([2, 1, 1, 0], (0, 1), 0.2, 1, 0.8))
-    print(ai.q)
 
 
 def test_best_future_reward(ai):
     print("\n--- Testing best_future_reward ---")
-    print(ai.best_future_reward([1,1,1,0]))
-    print(ai.best_future_reward([1,1,1,1]))
 
 
 def test_choose_action(ai):